The Agent Skills Directory

[External Downloads] (HIGH): The installation instructions require cloning a repository from https://github.com/DunbrackLab/IPSAE.git. This organization and repository are not on the 'Trusted External Sources' list, making the dependency unverifiable.
[Remote Code Execution] (HIGH): The skill documentation explicitly instructs users to execute code from the downloaded repository using python ipsae.py. This is a classic 'download then execute' pattern which allows for arbitrary code execution if the remote repository is compromised or malicious.
[Indirect Prompt Injection] (LOW): The skill provides a Python function (score_designs) that iterates over files in a directory and uses subprocess.run to execute the scoring script. While it does not use shell=True, it processes untrusted file system data (JSON and PDB files) which constitutes an ingestion point for external content. The capability is limited to local script execution.
[Command Execution] (LOW): The skill utilizes subprocess.run to call external Python scripts. This is part of the intended functionality but increases the attack surface if input file paths are not strictly controlled.

ipsae